0

I have a small time series data with 10 observations

Each observation is spaced at 20 days gap

For example, I have sales revenue from day 1, day 21, day 41, day 61 till day 221...

Now I would like to predict what will be the sales revenue on day 500th day amd 850th day?

How can we predict using short timeseries data (which has only 10 observations)?

Is there any specific technique that I should follow?

Can you guide me on how would you guys do this? What are some of the topics that I should learn to do this? Can someone please help me with list of steps on how to do this?

The Great
  • 1,380
  • 6
  • 18
  • Many similar Qs here, [look throgh this list](https://www.google.com/search?channel=fs&client=ubuntu&q=site%3Ahttps%3A%2F%2Fstats.stackexchange.com+predicting+short+time-series) – kjetil b halvorsen May 08 '21 at 15:47
  • 1
    You have (very little) data for days 1 to 221 and want to forecast for day 500 and day 850. That is, you are extrapolating out almost three times as far as the period your historical data covers is long. Be extremely careful about this, especially with only ten data points. If you fit any kind of trend, it will almost certainly be spurious, and the extrapolation will be catastrophic. My recommendation: use the historical average, and spend your time getting more and better data. – Stephan Kolassa May 10 '21 at 13:19
  • @StephanKolassa - by historical average, you mean for example, for forecasting 7th observation, compute the mean of 1st 6 observations, for forecasting 8th observation, compute the mean of 1st 7 observations, for forecasting 9th observation, compute the mean of 1st 8 observations and so on? Did I understand that right? – The Great May 10 '21 at 14:31
  • 1
    Exactly right! See also https://stats.stackexchange.com/q/124955/1352 – Stephan Kolassa May 10 '21 at 14:32
  • Another problem with short time series is we may not know whether there are any seasonality factors involved in it. Therefore, any assumptions/inference we make are nothing better than random. Is there any other problems that you can think of from short time series? – The Great May 10 '21 at 14:34

0 Answers0